Maturing monitoring agents into model based diagnostic agents for ground processing of the Space Shuttle and future exploration

NASA Kennedy Space Center deploys rule-based software agents to help monitor the Space Shuttle ground telemetry data. The agents recognize predefined measurement patterns and issue notifications to Shuttle Engineers when various events occur. Hundreds of rules for thousands of measurements have been written. Currently, these agents possess only shallow knowledge. They do not lend themselves to more complex tasks, such as autonomously diagnosing hardware failures in the ground support equipment. Processing and preparing a Space Shuttle for launch requires the verification of thousands of end items, such as pumps, valves, relays, and batteries. Diagnosing hardware problems within this extensive array of ground support equipment is a labor intensive task. It requires thousands of person hours for each processing flow of the Space Shuttle. Reasoning about the structure and behavior of the ground processing systems is ubiquitous in this real-time environment. Automating this diagnostic task would not only benefit today´s ground processing needs, but pave the path for infusing autonomy into future support systems that may operate on various non-terrestrial surfaces including lunar, Martian, and orbiting spaceports. When failures in surface support equipment occur in these remote locations, an army of humans-in-the-loop, as present in today´s Space Shuttle ground processing environment, will simply not be available. Thus, higher degrees of diagnostic automation becomes even more essential for remote surface support processing and operations. This paper presents an evolution path for maturing today´s monitoring agents into model based diagnostic agents that can benefit both the Space Shuttle and future Exploration vehicles. This will enable more automated inference of the location of end item failures. We discuss the identification of knowledge patterns within the existing Space Shuttle ground support domain, the migration of rule sets towards an XML like representation,- - and the introduction of an assumption based truth maintenance system to leverage this knowledge base